Yvonne Posted February 4, 2013 Posted February 4, 2013 There's a difference between being able to call that during a game and calling that after the result is known, and even then you have false positives and false negatives. Now, if we ever played enough games to have meaningful statistical power, then we'd be able to talk about the significance properly.
Sheikah Posted February 4, 2013 Posted February 4, 2013 Actually did call this to myself as I played the game and that's why I targeted him. Otherwise you're suggesting I targeted him at random, which I know to be untrue. I don't think I have any false positives on this particular point given I haven't seen him play like this before when not Mafia. True, I haven't exactly played a huge number of games, but I think I'd be a bad player if I hadn't picked up on some player trends by now. Interesting on the stats you mention; with a large enough measurable effect size the n number of samples (games) can be quite small. Not that anyone is going to be doing stats anytime soon, just putting it out there that you wouldn't really need many examples to form a significant conclusion. Our lab has managed less than 5 with some measurements.
Rummy Posted February 4, 2013 Posted February 4, 2013 WHO THE FUCK KILLED YVONNE?! Me. WHO SWITCHED OFF THE FUCKING POWER?! Aaand....Me. For actual comments; I think you did some really nice things with this @Mr\-Paul, who pretty much carried us, and I enjoyed our objective of trying to get all the electrics(we just had Peeps to go!) - how did Cube die last night, btw? We were hoping we might lynch Peeps with his help and steal the win last minute!
Cube Posted February 4, 2013 Posted February 4, 2013 how did Cube die last night, btw? We were hoping we might lynch Peeps with his help and steal the win last minute! I'm guessing he made me target myself. I chose to still attempt a kill, even though I knew this was likely.
Rummy Posted February 4, 2013 Posted February 4, 2013 Ah, of course. It was us who tried to re-direct Murray the night before to kill Peeps, but Peeps was redirecting him to Jimbob and I think Jay ensure it happened. We just came so close! Pretty happy how close considering the difficulties in actually taking folks out.
The Peeps Posted February 4, 2013 Posted February 4, 2013 I was trying to hold off revealing myself as electrical for as long as possible but you probably knew I was from night 1 anyway... Yeah I redirected Cube to himself in the last night phase.
Yvonne Posted February 4, 2013 Posted February 4, 2013 You can only call any level of significance with 4 results if they're all positive or all negative... This also ignores that people grow and adapt to the data as it's being made. In essence I'm saying relying on past behaviour is a recipe for getting fooled in the future, especially when people know that's what you're doing.
Cube Posted February 4, 2013 Posted February 4, 2013 I've seen similar "Cube is quiet like this when he's mafia" comments the last few times that I was part of the town.
The Peeps Posted February 4, 2013 Posted February 4, 2013 In my experience Cube has usually been more active when he's been town but you can't rely on that. There are too many other factors that can contribute to how active someone is. I thought you were active in this game Cube and you were even swaying me to your side after the lynch train. It's why I sent EEVIL to Jimbob and not you. In the end it just had to be you though :p
Cube Posted February 4, 2013 Posted February 4, 2013 It's why I sent EEVIL to Jimbob and not you. In the end it just had to be you though :p We decided to have him make the kill because I thought you were more likely to make me target EEVIL, which would have resulted in me reverse tracking him.
Sheikah Posted February 4, 2013 Posted February 4, 2013 You can only call any level of significance with 4 results if they're all positive or all negative... You do realise that 'statistical significance' is an arbitrary 95% confidence interval set a long time ago by a guy working for Guinness? Don't get too hung up about it. For a game like this I don't see any reason why you would demand significance. This also ignores that people grow and adapt to the data as it's being made.In essence I'm saying relying on past behaviour is a recipe for getting fooled in the future, especially when people know that's what you're doing. It appears I am being misunderstood. I'm not saying 'rely only on character analysis!' Obviously, you have to adapt yourself and use what information you have as well as sussing someone's behaviour. To claim that making some decisions based on how a player is behaving (like thinking they're behaving differently so choosing to target them) is a bad idea is the same as saying that you should only play poker based on the cards you have. It's silly. Sure, in both cases people can change the way they behave to fool people. But at the same time the answer isn't to bury your head in the sand and pretend judging people's behaviour against previous performance is a bad idea. You just have to find the right mix.
Yvonne Posted February 4, 2013 Posted February 4, 2013 Don't know where 5% came from... but anyway, it seems that we agree that no single strategy is a magic bullet for determining alignment.
EEVILMURRAY Posted February 5, 2013 Posted February 5, 2013 We decided to have him make the kill because I thought you were more likely to make me target EEVIL, which would have resulted in me reverse tracking him. OOOOH NOOOO. Don't say "we". I recommended we hold off making a kill that night for that very reason.
Mr_Odwin Posted February 5, 2013 Posted February 5, 2013 You do realise that 'statistical significance' is an arbitrary 95% confidence interval set a long time ago by a guy working for Guinness? That's the origin of the t-test. It's generally accepted that Fisher came up with p<0.05 as a cut-off. /leavesthread
Sheikah Posted February 5, 2013 Posted February 5, 2013 Ah I didn't know that came first. Still pretty arbitrary and annoying to scientists. :p
Recommended Posts